A study was mad e of data obtained over an 18·month period (July 1961to December 1962 on th e co mpari so n of atomic frequency s tandard s located in se ve n laboratories in the Un ited States, Europe, and Ca nada, using th e VLF signals of eBR (16 kc/s), Rugby, England, and NBA (18kc/s), Balboa, Canal Zone. Eac h laboratory observes th e accu mulated diffe ren ce in phase over a 24-hr period (the sa me for a ll laboratories, or nea rly so) be twee n its own sta ndard (e ith e r laboratory or commerciaUy co ns tru cted) and the receive d VLF signal. A stati sti cal analysis was des igned to se parate the obse rvation s at eac h laboratory into three components: (a) lon g-term mean diffe rences a mong the atomic standard s; (b) es timates or the standard deviations , a;, at eac h receiving stati o n; and (c) es timates of the tran s mitt e r standard deviations, 7. Each a; inc lud es rece iver flu ctuations, propaga tion effects pe rc uliar to th e pat h, and meas ure me nt unce rtainties; T in clud es the tran smitter flu c tuation s and propagation effects co mmon to all pat hs.Th e study shows that aLat each .receiver varied from a low of 0.39 X 10-1 0. units of fractional frequency (that is, 0.39 parts in 10 10 ) (GBR data) at LSRH to a high of 1.97 X 10-10 (GBR data) at NRC with an average for all s tations of 1.01 X 10-10 meas ured against eBR and 0.99 X 10-10 when meas ured against NBA. Also, the average T for eBR is 1.26 X 10-10 and for NBA is 0.68 X 10-10 • Fi na ll y, it is shown that: (1) the mea ns of the frequ e nc ie s of th e seven individual laboratories agreed with the grand mean of these seven laboratories to within ± 2 parts in 10 10 for the 18-month period , and (2) the laboratory-type standards agreed with th eir grand mean to within ± 1 part in 10 10